Why an Experience Plan Needs a Benchmark
An experience plan compares a risk’s observed losses with a reference amount. That reference is the plan’s expected losses: an estimate of the aggregate losses that would be expected for the risk over the stated experience period and coverage basis. Without a benchmark, an actual loss total has little context. A loss total of $40,000 could be high for one risk and low for another.
The comparison is meaningful only when the benchmark and observed losses refer to the same risk, coverage, and experience period, and use compatible loss valuation and maturity. Those bases should be established before the comparison is made. The expected amount is a plan benchmark, not a promise that the risk will incur exactly that amount and not, by itself, a premium.
In a simple comparison, let \(E\) denote expected losses and \(A\) denote actual losses. The dollar difference \(A-E\) measures how far observed losses are above or below the benchmark. The ratio \(A/E\), when \(E>0\), expresses actual losses as a proportion of expected losses. A ratio above 1 indicates actual losses exceeded the benchmark; a ratio below 1 indicates they were lower. Neither measure alone specifies the premium change: a plan may apply additional rules, limits, or adjustments.
Expected Loss Dollars Are Not Expected Claim Counts
Claim count and loss amount describe different aspects of experience. A risk can have more claims than expected but lower losses than expected if the claims are relatively small. Conversely, fewer claims than expected can produce high aggregate losses if one or more claims are severe. An expected count can be fractional: it is a probability-weighted average across possible outcomes, not a prediction that a fraction of a claim will occur.
A standard frequency-severity model helps show how expected claim counts relate to expected losses. Let \(N\) be the claim count, and let \(X_j\) be the loss amount for claim \(j\). Aggregate losses are \(L=\sum_{j=1}^{N}X_j\). If claim severities have a common finite mean \(\mu_X\), and the severities are independent of the claim count, then expected aggregate losses equal expected claim count times expected severity.
The first equality is the law of iterated expectations. Conditional on \(N=n\), the expected total is \(n\mu_X\), because it is the sum of \(n\) claim amounts, each with mean \(\mu_X\). Taking the expectation over the possible values of \(N\) gives \(\mu_X E[N]\). The claim-count and severity assumptions matter: if severity depends on the number of claims or the claim population is not represented by a common mean, the simple product need not apply. In any case, the product estimates dollars, while \(E[N]\) alone estimates a count.
Worked Example: Separating Count and Dollar Comparisons
A hypothetical equipment policy has expected claim count 8.0 and expected severity $5,000 per claim for its experience period. Assume \(N\sim\operatorname{Poisson}(8)\), claim severities are fixed at $5,000 and independent of \(N\), and the plan uses the resulting expected aggregate as its benchmark. The observed period has 10 claims with total actual losses of $47,500. Find the expected losses, the count comparison, and the actual-to-expected loss ratio. Assume no discounting or interest adjustment and no mortality component.
Method and benchmark. The Poisson mean is \(E[N]=8\). With fixed severity \(X_j=\$5{,}000\), the expected aggregate is
The plan’s expected-loss benchmark is therefore $40,000. The observed claim count is 10, which is 2 claims above the expected count; its actual-to-expected count ratio is \(10/8=1.25\).
Dollar comparison. Actual losses are $47,500, so the dollar excess is \(\$47{,}500-\$40{,}000=\$7{,}500\). The loss ratio is \(\$47{,}500/\$40{,}000=1.1875\), or 118.75%.
As a check, observed average severity is \(\$47{,}500/10=\$4{,}750\). The actual-to-expected severity ratio is \(4{,}750/5{,}000=0.95\), and \(1.25\times0.95=1.1875\), agreeing with the aggregate loss ratio. The common mistake is to treat the 25% excess in claim count as a 25% excess in loss dollars; the observed lower average severity means the dollar excess is 18.75%, not 25%.
Expected Losses Are Not Prospective Premium
A prospective premium is the amount charged for future insurance coverage under the applicable rating and expense structure. Expected losses are a loss-cost benchmark for a specified basis and period. A premium may include provisions for expenses, taxes, contingencies, or other components, and may be affected by plan rules. Thus, a loss benchmark cannot be relabelled as premium, and a comparison of actual losses with expected losses does not by itself determine the future premium.
The distinction also prevents a common timing error. Expected losses are attached to the experience period and coverage basis used by the plan. A prospective premium applies to future coverage. Even if both amounts are expressed in dollars, they answer different questions: one benchmarks past experience; the other prices future insurance.
Worked Example: Keeping Benchmark and Premium Separate
A hypothetical contractor’s plan assigns expected losses of $42,000 for the stated experience period and coverage. Actual losses on the same basis are $49,000. For a separate prospective term, suppose the quoted premium is constructed from $42,000 of expected loss provision, $8,400 of expense provision, $2,100 of risk and profit provision, and $1,500 of taxes. Assume these are additive components, with no other adjustments, and that the stated amounts already use the applicable valuation basis. No interest discounting or mortality assumption applies, and no probability distribution is needed because the amounts are supplied.
Experience comparison. Actual losses exceed expected losses by \(\$49{,}000-\$42{,}000=\$7{,}000\). The actual-to-expected loss ratio is \(\$49{,}000/\$42{,}000=1.1667\), or 116.67% when expressed as a percentage to two decimals.
Prospective premium. Add the stated components:
The benchmark remains $42,000; the prospective premium is $54,000. The fact that the premium includes an expected-loss provision of $42,000 does not make the two totals interchangeable. A common mistake is to use $54,000 as the denominator in the experience loss ratio. That would mix premium with the loss benchmark; the correct denominator is $42,000.
Combining Benchmarks Without Losing Their Meaning
If a plan provides expected-loss benchmarks for several distinct components on compatible bases, the total benchmark is their sum. Actual losses must be aggregated over the same components before comparing totals. Component-level comparisons can still be informative: a favorable result in one component can offset an unfavorable result in another when only the total is considered.
This aggregation does not make the components interchangeable. The total answers how the combined experience compares with the combined benchmark. It does not show which component drove the result. If component-level results matter to the plan, calculate and report them as well as the aggregate.
Worked Example: Component Results and the Aggregate Benchmark
A hypothetical small-business plan tracks two coverages over the same experience period. For Coverage A, expected losses are $18,400 and actual losses are $21,000. For Coverage B, expected losses are $31,600 and actual losses are $28,800. All amounts are on matching coverage, valuation, and maturity bases. Treat the figures as supplied plan inputs; no interest, mortality, or probability distribution is needed. Calculate the component and total comparisons.
Coverage A. Its excess is \(\$21{,}000-\$18{,}400=\$2{,}600\). Its ratio is \(\$21{,}000/\$18{,}400=1.1413\), or 114.13%.
Coverage B. Its excess is \(\$28{,}800-\$31{,}600=-\$2{,}800\), so actual losses are $2,800 below expected. Its ratio is \(\$28{,}800/\$31{,}600=0.9114\), or 91.14%.
Combined result. Total expected losses are \(\$18{,}400+\$31{,}600=\$50{,}000\), and total actual losses are \(\$21{,}000+\$28{,}800=\$49{,}800\). The combined difference is \(\$49{,}800-\$50{,}000=-\$200\), and the combined ratio is \(\$49{,}800/\$50{,}000=0.9960\), or 99.60%.
The check is that component differences sum to the total difference: \(\$2{,}600-\$2{,}800=-\$200\). The overall result is slightly below expected even though Coverage A is substantially above its own benchmark. The common mistake is to average the two component ratios equally; they have different expected-loss amounts. Aggregating dollars first gives the correct total ratio.
Interpretation and Common Traps
Expected losses are estimates, not guaranteed outcomes. Random variation means actual losses can depart from the benchmark even when the benchmark is appropriate. A high actual-to-expected ratio describes the observed relationship; by itself, it does not establish why the result occurred or specify how a plan should respond. Interpretation also depends on the quality and consistency of the inputs.
Before calculating a comparison, check that the expected and actual amounts use the same coverage and time basis. Confirm that the observed losses are valued as the plan requires. Then keep the units straight: claim counts are numbers of claims, whereas expected and actual losses are dollars. Finally, keep premium separate from the loss benchmark.
- Using expected count as the loss benchmark. A count is not a dollar amount; a severity assumption or a separately provided expected-loss amount is needed to compare losses.
- Calling actual losses “expected.” Actual losses are the observed realization; expected losses are the benchmark estimate.
- Using premium as the denominator. Unless the plan explicitly defines a different comparison, an actual-to-expected loss ratio uses expected losses, not prospective premium.
- Combining mismatched amounts. A valid comparison requires compatible periods, coverages, and valuation bases.
Check Your Understanding
Work each comparison on the stated basis, then check the final answer.
- A hypothetical fleet has expected losses of $32,000 and actual losses of $28,800 for the same period and coverage. Find the dollar difference and actual-to-expected loss ratio. (Answer: actual is $3,200 below expected; ratio \(0.9000\), or 90.00%)
- A hypothetical warranty portfolio has \(N\sim\operatorname{Poisson}(3.2)\), with independent fixed claim severity of $4,500. Find expected losses. If 4 claims produce actual losses of $19,200, find the actual-to-expected loss ratio. (Answer: expected losses $14,400; ratio \(1.3333\), or 133.33%)
- For two compatible coverages, expected losses are $12,500 and $7,500; actual losses are $15,000 and $9,000, respectively. Find each component ratio and the combined ratio. (Answer: component ratios 1.2000 and 1.2000; combined ratio \(24{,}000/20{,}000=1.2000\))
- A hypothetical account has expected losses of $48,000 and actual losses of $60,000. Its prospective premium is $75,000. Calculate the experience loss ratio, and state whether the premium is the correct denominator. (Answer: \(60{,}000/48{,}000=1.2500\), or 125.00%; no, use expected losses, not premium)
- Expected claim count is 5, with independent claim severity having mean $6,000. Actual experience consists of three claims of $4,000, $7,000, and $13,000. Compare the actual-to-expected count ratio with the actual-to-expected loss ratio. (Answer: expected losses $30,000; actual losses $24,000; count ratio \(3/5=0.6000\), loss ratio \(24{,}000/30{,}000=0.8000\))